Levels of Measurement: Nominal, Ordinal, Interval, and Ratio

Nominal

  • Definition: Nominal (categorical) level classifies data into categories with no intrinsic order or quantitative meaning.
  • Key idea: There is no meaningful math that can be performed on nominal data aside from counting frequencies.
  • What’s not meaningful: ordering, adding, subtracting, or any arithmetic operations.
  • Example from transcript: hair color (e.g., black, blonde). Observations can be labeled as 'black hair' or 'blonde hair' but you cannot say one category is greater or smaller or perform arithmetic on them.
  • Data type: Qualitative; categories are labels rather than numbers.
  • Descriptive statistics: use frequencies and modes; visualize with bar charts or pie charts; suitable for chi-square tests of association when comparing groups.
  • Connections to practice: choosing the right tests and summaries depends on level of measurement; nominal data support non-parametric methods and categorical analyses rather than arithmetic means.
  • Practical implication: misapplying arithmetic (e.g., computing averages) to nominal data yields meaningless results.

Ordinal

  • Definition: Ordinal level provides categories with a natural, meaningful order, but the intervals between adjacent categories are not assumed to be equal.
  • Key idea: You can rank or order observations, but you should not assume equal spacing between ranks.
  • Example from transcript: academic grades A, B, C, D have a natural ordering (A > B > C > D).
  • What’s meaningful: ordering/ranking; differences between levels are not necessarily uniform.
  • What’s not always meaningful: arithmetic operations like addition or subtraction between categories; calculating a precise mean of ordinal categories is generally not appropriate.
  • Descriptive statistics: medians and modes are common; non-parametric tests (e.g., Mann-Whitney U, Kruskal-Wallis) are often used for inference.
  • Connections to practice: ordinal data arise in satisfaction ratings, class ranks, Likert-type scales; understanding scale informs which summary statistics and tests are valid.
  • Practical implication: treat with appropriate non-parametric methods when making inferences; avoid assuming equal intervals.

Interval

  • Definition: Interval level provides numeric values with equal intervals between points, but has no true zero point.
  • Key idea: Equal spacing between values is meaningful; differences (x − y) are interpretable.
  • Important caveat: Zero is arbitrary and does not indicate absence of the quantity.
  • Example from transcript: temperature scales (e.g., 0°C or 0°F) illustrate that zero does not entail 'no temperature'.
  • What’s meaningful: addition and subtraction of values are meaningful; differences are interpretable; you can compute measures like the mean and standard deviation.
  • What’s not meaningful: meaningful ratios are not generally supported because zero is not a true absence (e.g., 20°C is not twice as hot as 10°C).
  • Descriptive statistics: mean, median, standard deviation, and other parametric-style summaries are commonly used; confidence intervals and parametric tests are often appropriate when assumptions hold.
  • Formulas (typical interval-scale statistics):
    • Mean: ar{x} = rac{1}{n} ext{\sum}{i=1}^n xi
    • Variance/Standard deviation: $$ ext{Var}(X) = rac{1}{n} ext{\sum}{i=1}^n (xi - ar{x})^2,
      \ ext{SD} = \